Morphological Effects in SnO 2 Chemiresistors for Ethanol Detection: A Systematic Statistical Analysis of Results Published in the Last 5 Years

: SnO 2 is one of the most studied materials in gas sensing. Among the many strategies adopted to optimize its sensing properties, the ﬁne tuning of the morphology in nanoparticles, nanowires, and nanosheets, as well as their eventual hierarchical organization, has become an active ﬁeld of research. In this work, results published in the literature over the last ﬁve years are systematically analyzed focusing on response intensities recorded with chemiresistors based on pure SnO 2 for ethanol detection in dry air. Results indicate that no morphology clearly outperforms others, while a few individual sensors emerge as remarkable outliers with respect to the whole dataset.


Introduction
Chemiresistors based on semiconducting metal oxides are among the most popular gas sensing devices.Their success comes from their high sensitivity to a broad range of chemicals, their reduced size and power consumption, and their suitability for mass production at relatively reduced costs.To optimize the sensing layer, the fine control of the morphology, both at the level of individual nanostructures and at the level of their hierarchical assembly, has been reported as very effective [1,2].
In this work, with the aim to have a more general and reliable picture of the state of the art, results published in the literature in the last five years are systematically analyzed, focusing on response intensities recorded with chemiresistors based on pure SnO 2 for ethanol detection in dry air, as the case example.In particular, we chose to focus on SnO 2 because it is the most studied material among semiconducting metal oxides.Similarly, we chose ethanol as target gas because it is widely used as a test gas for the development of innovative materials (morphologies) and it is a key component in many applications [3].

Materials and Methods
This work considers the responses to ethanol reported for chemiresistors based on pure SnO 2 in the period from January 2015 to July 2020.In order to have a common background between all the considered responses, only dry air tests have been taken into account.
The morphology of the SnO 2 layer is described at two different levels: at the level of individual nanostructures and the level of their eventual hierarchical assembly.
Concerning the shape of individual crystallites composing the sensing layer, it has been categorized as follows: • Nanorods: elongated nanostructures with a high aspect-ratio, and surfaces identified by well-defined crystalline planes;

Results
As an example of the shape of elementary nanostructures widely investigated in the literature, Figure 1 reports the SEM images for two SnO 2 layers composed by a disordered network of nanowires (Figure 1a), and by a disordered network of nanoparticles (Figure 1b) [1].Therefore, some nanoparticles are distributed over the substrate individually, while others are distributed in µm-sized grains as a consequence of aggregation often observed in nanoparticle-based layers [1].

•
Nanorods: elongated nanostructures with a high aspect-ratio, and surfaces identified by well-defined crystalline planes; • Nanoparticles: spherical nanostructures, such as those used in thick films; • Nanosheets: thin nanostructures extending in two dimensions.

Results
As an example of the shape of elementary nanostructures widely investigated in the literature, Figure 1 reports the SEM images for two SnO2 layers composed by a disordered network of nanowires (Figure 1a), and by a disordered network of nanoparticles (Figure 1b) [1].Therefore, some nanoparticles are distributed over the substrate individually, while others are distributed in μm-sized grains as a consequence of aggregation often observed in nanoparticle-based layers [1].Boxplots resuming the responses to 10 ppm and to 300 ppm of ethanol reported in literature are shown in Figure 2a and 2b, respectively, grouping the results by nanostructure morphologies, namely nanorods, nanoparticles, and nanosheets.
The statistical parameters describing these distributions are reported in Tables 1 and  2 for data shown in Figure 2a and 2b, respectively.
Statistical parameters reported in these tables are: the number of samples considered in each category (morphology of elementary nanostructures); the number of outliers identified for each category; the values of the 1st, 2nd, and 3rd quartiles (Q1, Q2, and Q3) of the response amplitude Ggas/Gair; and the values of the upper and lower whiskers.The pvalue of the median test comparing the median response of morphologies two by two are also reported in order to have a statistical check about the similarity and dissimilarity between median responses of the different morphologies.The statistical parameters describing these distributions are reported in Tables 1 and 2 for data shown in Figure 2a,b, respectively.Statistical parameters reported in these tables are: the number of samples considered in each category (morphology of elementary nanostructures); the number of outliers identified for each category; the values of the 1st, 2nd, and 3rd quartiles (Q1, Q2, and Q3) of the response amplitude G gas /G air ; and the values of the upper and lower whiskers.The pvalue of the median test comparing the median response of morphologies two by two are also reported in order to have a statistical check about the similarity and dissimilarity between median responses of the different morphologies.

Discussion
The distributions of the response intensities shown in Figure 2 depend on the gas concentration.This is partially due to the fact that different authors often tested their sensors against different ethanol concentration so there is no a complete overlap between concentration used in different articles.In other words, the sensors whose response is shown in Figure 2a are not exactly the same sensors whose response is shown in Figure 2b.Nonetheless, despite these differences, a common feature is that no morphology clearly performs better than other morphologies.Median tests reported in Tables 1 and 2 feature a p-value that is larger than 0.05 in all situations.This means that there is no clear evidence to reject the null hypothesis, i.e., there is no clear evidence to reject the hypothesis that the couple of morphologies under the test are not distinguishable.The same is observed for other concentrations and also considers the eventual hierarchical organization of the individual nanostructures into assemblies, such as hollow spheres, fibers, hollow fibers, etc. [46].On the other hand, some materials emerge as outliers with respect to all morphologies.In Figure 2a, there are five outliers: four are the responses from layers composed by nanoparticles, namely [4][5][6][7] with response intensities of about 236, 50, 49, and 50 (to 10 ppm of ethanol), and one composed by nanosheets [33] featuring a response G gas /G air ≈ 50.As a reference, the median responses to this ethanol concentration are around 4.55, 2.3, and 10 for nanoparticles, nanorods, and nanosheets, respectively.Concerning the concentration of 300 ppm, four outliers emerges: the nanoparticles synthesized by [45], and two types of nanorods and the nanosheets developed by [38].These materials feature responses of about 2000, 4070, 1609, and 495, compared with the median responses of 71, 52, and 38 for nanosheets, nanorods, and nanoparticles, respectively.
These results are arguably due to the longer tradition of the synthesis of nanoparticles with respect to those of nanowires and nanosheets.Such a longer experience may reasonably imply a more developed capability to effectively combine the many parameters underlying the sensing mechanism, which may counterbalance the advantages arising from the fine morphological tuning inherent in the more recent nanostructures.

Figure 1 .
Figure 1.Examples of two different morphologies investigated in the literature for SnO2-based chemiresistors.(a) Film composed by a disordered network of SnO2 nanowires; (b) film composed by a disordered network of SnO2 nanoparticles, which are distributed either individually or in μmsized aggregates.Reprinted from [1].

Figure 1 .Figure 2 .
Figure 1.Examples of two different morphologies investigated in the literature for SnO 2 -based chemiresistors.(a) Film composed by a disordered network of SnO 2 nanowires; (b) film composed by a disordered network of SnO 2 nanoparticles, which are distributed either individually or in µm-sized aggregates.Reprinted from [1].Boxplots resuming the responses to 10 ppm and to 300 ppm of ethanol reported in literature are shown in Figure2a,b, respectively, grouping the results by nanostructure morphologies, namely nanorods, nanoparticles, and nanosheets.Chem.Proc.2021, 5, 75 3 of 4

Figure 2 .
Figure 2. Boxplots resuming the statistics of the response intensities of SnO 2 chemiresistors grouped by crystallite shape.(a) Statistics recorded vs. 10 ppm of ethanol; (b) statistics recorded vs. 300 ppm of ethanol.

Table 1 .
Statistics of data shown in Figure1a(responses to 10 ppm of ethanol).

Table 1 .
Statistics of data shown in Figure1a(responses to 10 ppm of ethanol).

Table 2 .
Statistics of data shown in Figure1b(responses to 300 ppm of ethanol).